<?xml version="1.0" standalone="yes"?>
<Paper uid="W94-0321">
  <Title>Recognizing Digressive Questions During Interactive Generation *</Title>
  <Section position="9" start_page="187" end_page="187" type="concl">
    <SectionTitle>
7 The System Implementation
</SectionTitle>
    <Paragraph position="0"> In the knowledge base, the several sources of information that the system needs to analyze and plan the discourse are all represented uniformly using the Semantic Network Processing System (SNePS) \[17\]. This includes knowledge of the text plan operators, the domain plans, entities in the domain, the user model, the discourse plan executed so far, and rules for reasoning about all of the above. The SNePS Actor models a cognitive agent operating in a single-agent world. It integrates inference and acting by representing beliefs, plans, and acts as structured intensional entities in the network formalism. Because the agent's world and planning knowledge is represented uniformly, the agent can discuss, reason about, formulate, and also execute its plans.</Paragraph>
    <Paragraph position="1"> Based on the TOUR model \[8\], the various driving routes that my system can discuss are represented as preconstructed plans that are composed of two types of actions: going and turning. The domain plans are represented at various levels of detail and, as conceptual entities, can have properties. Whenever the system reasons about the domain, the reasoning that leads to deductions is recorded in the knowledge base along with the deductions themselves and is available as content for explanatiol,s.</Paragraph>
  </Section>
class="xml-element"></Paper>